Capturing and processing camera and sensors data and recognizing various forms to determine a set of robotic actions is conceptually easy. However, Amazon has challenged the challenge of selecting and choosing stains in a robotic way and 28 teams from the whole world rose to competition.
Perception does not only concern cameras and sensors. The software must convert the data and deduce into what it “sees”. In the case of Amazon selection challenge held last week at International IEEE conference on robotics and automation (ICRA)The robot of each team had to choose from a list of races of consumption elements of variable shapes and sizes – pencils, toys, tennis balls, cookies and cereal boxes – which were stored randomly on The shelves, then place their selected items in a tank. They could use any robot, mobile or not, and any tools or end of arms entry or tools to accomplish the task.
It is difficult for robots using sensors to identify and locate objects that can be confused by plastic packaging in the platform or storage area. Rodney Brooks, Irobot, MIT and Rethink Robotics Fame, often speaks of an aspiration objective on the scale of industry concerning the perception of robotics: “If we could only provide the visual capacities of A 2 year old child, robots would quickly get much better. This is what this competition is used for.
The software must first identify the element to choose, then find the best way to catch it and move it outside the storage area. Amazon, with its acquisition of Kiva Systems, has mastered the goods to the picker / Packer and now wants to automate the remaining process of selection of correct products of the shelves and to place them in the packaging box, hence their selection challenge D ‘Amazon.
The three best winners were the teams of the Technical University (Tu) – Berlin, with 148 points; Mit in 2nd place with 88 points; And the finisher of 3rd place (Oakland U and Datasepeed) which only obtained 35 points. The teams were noted on the number of elements selected, selected and placed.
Many commercial companies with proprietary software for this type of application (such as Universal robotics and their neocortex vision systemSilicon Valley startup Recover robotics who was first of all their new Recovery and freight system at the same ICRA conference, and other startups that are less far from Harvard’s spin-out Right robotics) chose not to come in because the terms of the challenge understood that the software was opened.
The winning group, Tu-Berlin RBO teamChose a mobile manipulator as an approach to the competition solution. They have written their own vision system software and will soon work on an article on the subject. They used a Barrett Wam Arm Because it was the most flexible device to provide a combination of reactivity, geometry, agility, dexterity and foundability in conjunction with their end -of -arm tool. They used an increased vacuum cleaner tool with a taster and vacuum cleaner to supply suction. For a base, they used an old mobile platform for nomadic technologies that they have improved to meet the needs of the competition. (Nomadic no longer exists. It was acquired by 3COM in 2000.) You photo on the right shows the view of the robot camera and the random placement of the elements in the cubby holes.
Noriko Takiguchi, a Japanese journalist for ROBONEWS.NETWho was in the contest, observed the team you and said that their approach was a more position on the torque of the arm and the mobile base, therefore, they had good control of the couple which allowed them to ‘Have flexibility in the choice of the place to place the cup aspiration and how much suction to apply.
The RBO team received a first place for $ 20,000 plus travel costs for equipment and team members.